Detecting pulmonary Coccidioidomycosis with deep convolutional neural networks

Artificial intelligence in veterinary diagnostic imaging: A literature review

“…Ott et al 43 . developed five different NNs based on Inception, MobileNet, ResNet, VGG, and a shallow four‐layer network to specifically detect pulmonary lesions caused by coccidiomycosis in lateral and ventrodorsal radiographic projections.…”

“…The ability of the ResNet model to locate radiographic changes related to coccidiomycosis was also explored to ensure the NN had reached its conclusions based on true disease‐related changes rather than other features in the images (such as laterality markers). Class activation maps were generated to demonstrate the spatial regions within radiographs that the network weighed highly when performing its classification 43 . The high performance of the network across multiple radiographic machines and thoracic projections is encouraging, however, the NNs were not asked to differentiate coccidiomycosis lesions from other pulmonary changes (notably, other granulomatous or neoplastic etiologies), an important consideration prior to clinical implementation of such an algorithm.…”

“…Three papers focus exclusively on the cardiac silhouette. [38][39][40] Four papers address pulmonary parenchymal changes: three using radiography and one using CT. [41][42][43][44] Five groups applied AI for comprehensive evaluation of canine and/or feline thoracic radiographs, including evaluation of pulmonary, cardiovascular, pleural, mediastinal and musculoskeletal structures, and one focused on the pleural space. [45][46][47][48][49] equivalent, and agreement between the two methods was 85%.…”

“…42 The higher accuracy achieved for alveolar patterns in this study is similar to human reader performance, in which identification of an alveolar pattern is agreed upon more often amongst radiologists than interstitial or bronchial patterns. 51 Ott et al 43 41 These results are reflective of clinical reality in which normal and abnormal lungs are relatively easy to differentiate, while determining the underlying cause for pulmonary parenchymal changes based on radiographs is often challenging, necessitating multiple differentials.…”

“…43 The high performance of the network across multiple radiographic machines and thoracic projections is encouraging, however, the NNs were not asked to differentiate coccidiomycosis lesions from other pulmonary changes (notably, other granulomatous or neoplastic etiologies), an important consideration prior to clinical implementation of such an algorithm.…”